Inverse estimation of the emission of radioactive materials from Fukushima Taichu Y. Tanaka, Takashi Maki, Mizuo Kajino, and Tsuyoshi T. Sekiyama Meteorological Research Institute, Japan Meteorological Agency 16 May 2012, ICAP workshop@esa/esrin, Frascati
Outline Introduction Global simulation of radioactive materials from Fukushima Preliminary results of the inverse estimation of the emission flux of 137 Cs from the accident.
On 11 March 2011, 2011 Tohoku earthquake occurred: A Huge earthquake with Mw9.0 occurred off the Sanriku, approximately 180km off the Pacific coast of Japan, and a large Tsunami attacked the Pacific coast of Tohoku (North eastern region) and Northern Kanto regions in Japan. More than 19,000 people have been missing or passed away. (From Wikipedia)
At Fukushima prefecture, The Tsunami flooded the Fukushima Dai ichi nuclear power plant (Fukushima NPP1), and caused a total loss of electricity which leads to the meltdown of the reactors. Huge amount of radioactive materials are released to the environment. (From TEPCO report)
Motivation of our study This disastrous accident of nuclear power plant should be investigated and documented as possible in detail for the future generation. The amounts of released radioactive materials are essential data. We should improve the atmospheric tracer models and analysis technique by evaluating the emission, transport and deposition of this event.
Estimated amount of released 137 Cs Japan Atomic Energy Agency (Chino et al., 2011) 12.6 PBq (3/12 4/5) JAEA, revised 9.1 PBq (3/11 4/5) JNES report 15 PBq* (Initially, 6.1PBq) Stohl et al. (2012,ACP) 36.6 PBq(20.1 53.1PBq; 3/10 4/20) Winiarek et al. (2012,JGR) 10 19 PBq IRSN, France 10 PBq ( 3/22) ZAMG 66 PBq (the first four days) More than factor of 5 difference! *Report of Japanese Government to the IAEA Ministerial Conference on Nuclear Safety The Accident at TEPCO's Fukushima Nuclear Power Stations, June, 2011 http://www.irsn.fr/en/news/documents/irsn_fukushima radioactivity released assessment EN.pdf
Simulation of 222 Rn and 210 Pb 222 Rn 210 Pb DJF (Winter in NH) JJA (Summer in NH) Above: 222 Rn, below: 210 Pb, surface concentration
Simulation of 85 Kr from nuclear reprocessing plant Simulated surface concentration of 85 Kr from 1955 to 2000.
Comparison with observed concentration of 85 Kr at MRI/JMA, Tsukuba Observation (Igarashi et al. 2000) Simulated with MASINGAR France UK Russia Japan USA Others
Simulation of the radioactive materials from the Fukushima accident
Brief description of the global model We used our global model called MASINGAR (Model of Aerosol Species in the Global Atmosphere), which calculates the distribution of aerosol species. MASINGAR has been used for the Asian dust information by JMA, and climate change experiments by Meteorological Research Institute. Here we will show the results of the developmental version of the model (MASINGAR mk 2). 3 dimensional image of transported dust simulation using MASINGAR http://www.jma.go.jp/jp/kosa/
Experimental settings Included radionuclides: 6 species I 131, Cs 137, Cs 134, Te 132, I 132, Xe 133 Xe 133 is treated as non reacting gas with no dry and wet depositions. Other species are assumed to be attached to aerosols (Lognormal size distribution with r n =0.07μm, σ=2.0, and hydrophilic) Model resolution: Horizontal TL319 (0.56 o, approx.60km), vertical 40 layers (ground 0.4hPa) Atmospheric dynamical model: JMA/MRI unified general circulation model (MRI AGCM3) Horizontal wind components are nudged using JMA global analysis (GANAL).
Estimated emissions from Fukushima Dai ichi nuclear power plant Source: 131 I, 137 Cs, 134 Cs, 132 Te : JAEA (Chino, personal communication) 133 Xe: Stohl et al. (2011) For 137 Cs, estimated emission by Stohl et al. (2011) is used for a sensitivity study.
Japan Atomic Energy Agency (JAEA) estimated emission amount of radio nuclei by comparing model simulation results and observation data (Chino et al., 2011) Prior studies (1)
Prior studies (2) Norwegian Institute for Air Research(NILU) estimated emission amount of radio nuclei by their Lagrangian transport model and inverse model (Stohl et al., 2011). Emission amount (upper) and their uncertainty Available observation sites (43 sites in globe, 5 sites in Japan)
Simulated Cs 137
Comparison of simulated results with observations. Currently, we obtained the observed concentrations of radioactive materials from: MRI, Tsukuba (Y. Igarashi) The Comprehensive Nuclear Test Ban Treaty Organization (CTBTO) Ring of 5 (Ro5, Masson et al. 2011) Published references (Berkeley, Bowyer et al. 2011; Hoffman et al. 2011; Hsu et al. 2011)
Comparison of observed and simulated time series at Tsukuba, Japan Time series of observed radionuclides on aerosols at Tsukuba, Japan (Igarashi et al., MRI) Simulated concentrations of I 131, Cs 137, Cs 134, and Te 132
Comparison of observed and simulated 133 Xe Spitsbergen, Norway Charlottesville, USA Takasaki, Japan Oahu, Hawaii, USA
Comparison of observed and simulated 137 Cs Spitsbergen, Norway Charlottesville, USA Takasaki, Japan Oahu, Hawaii, USA
Total deposition of 137 Cs (JAEA emission inventory) Deposited over land: 29.9% (2.0 PBq) Deposited over ocean: 70.1% (4.8 PBq) Most of the 137 Cs is deposited over Northern Pacific
Total deposition of 137 Cs (Emission estimated by Stohl et al. 2011, ACPD)) Emission : Stohl et al. (2011) Deposited over land: Deposited over ocean: 17.6% (6.1 PBq) 82.4% (28.8 PBq)
Comparison of 134 Cs and 137 Cs total deposition with aircraft observation Observed 134 Cs+ 137 Cs deposition by aircraft measurement (MEXT, 11 Nov. 2011) Simulated deposition with JAEA emission inventory Simulated deposition with emission estimated by Stohl et al. (2011, ACPD)
The emission flux estimation system by inverse model 1. Concept of inverse model 2. Estimation of dust flux emission 3. Application to an accident of Fukushimadai ichi nuclear power plant Principal Investigator: Takashi Maki
The concept of Inversion Aim To estimate initial nuclei emission flux by inverse model Assumption Observation (y) is determined by linear combination of emission flux (x) and Transport model processes (A). Features We obtain initial emission flux not only by observation (y) and transport model (A) but also prior information (x p ). The precision of the estimation depends on a quality of prior information. S T 1 T 1 x Ax y C Ax y x x C x x y We obtain x which minimizes the cost function S(x). p x p
The Bayesian Synthesis Inversion n mn m n m x x x A A A A A A A y y y 2 1 1 22 21 1 12 11 2 1 Observation results Transport matrixes Flux at grid point p x T p y T S x x C x x y x C y x x 1 1 A A Cost function Observation error Prior flux error We determine flux x which minimize cost function S(x). x p : prior flux Observation y x C y x A A 1 y T Shows the difference between model and observations p x T p x x C x x 1 Shows the difference between results and prior information
CO 2 flux analysis by inversion Dec. 2000 Dec. 2012 Greenhouse gas observations are made at about 150 stations worldwide. Global distributions of greenhouse gases are obtained from such limited observations using numerical models that simulate atmospheric transport and estimate CO 2 fluxes inversely from observations (Maki et al., 2010). http://ds.data.jma.go.jp/ghg/kanshi/info_kanshi_e.html
Dust emission analysis by inversion Observation network of PM10 We could obtain more realistic dust concentration time series by inverse modeling. We need more observation data near dust source area (Maki et al., 2011).
Dust Flux estimation results Table 2. Total estimated flux around the dust source area during the analysis period (26 Mar. 2007 3 Apr. 2007). Unit is Tg. Prior flux uncertainty Exp. ADORC Exp. KOREA Exp. JK Exp. ALL MASINGAR 0.2 (20%) 20.4 23.6 23.8 23.4 0.5 (50%) 21.6 32.3 30.9 32.6 1.0 (100%) 22.0 42.2 37.1 42.7 19.5 * We assume the dust source area to be a box region (Fig. 1) containing 44 grid cells in MASINGAR from 94 degrees east to 125 degrees east and from 36 degrees north to 47 degrees north. Exp. JK The estimated total dust flux amount are sensitive to observational data and their uncertainty. We need more observation data near dust source area (Maki et al., 2011).
Application to Fukushima accident Huge amount of radio nuclei were emitted from Fukushima Dai ichi nuclear power plant. Some institute forecasted spread of radio nuclei aerosol, but these information were not used to evacuation activity efficiently. Fukushima power plant (TEPCO HP) One of the main reason is that we could not know a amount, time series of radio nuclei emission. This hazardous accident has a feature that horizontal source area is known. We should use such information for emission estimation.
The feature of this research JAEA NILU MRI Number of observation 17 (17) 43 (5) 49 (49) Transport model WSPEEDI FLEXPART MRI PM/r MASINGAR mk 2 Model type Lagrange Lagrange Euler Meteorological Input JMA GSM (0.25 x 0.2 deg) ECMWF (0.18 deg) #GFS (0.5 deg) MANAL (0.05 deg) JCDAS (1.25 deg) Flux estimation Peak comparison Inversion Inversion Time resolution variable hours 6(3) hours output Daily # in second row shows observations in Japan mainland. We consider it possible to analyze detailed estimation flux by using Japanese observation.
Preliminary result of our Inverse analysis with the global model Observation: 50 sites observation data from CTBT, Ro5 and so on. Model: MASINGAR mk2 (Tanaka et al.) Prior Information: Estimation flux by Chino et al. (JAEA) Total Cs 137 flux: 9 PBq 13 PBq (increase) Preliminary estimated emission fluxes.
Differences using other first guess emission Here, we also used estimation flux by Stohl et al. (2012) for the inverse analysis. Total Cs 137 flux: 29 PBq 24 PBq ( decrease) Preliminary estimated emission fluxes.
In our inversion analysis, Estimated emission amount from Hydrogen explosion of the Unit 3 on 14 March was increased, with both first guesses. Both first guesses underestimated the release from the Unit 3 explosion? We should examine the height of the explosion. Explosion of the Unit 3 of Fukushima NPP1. (from TV news)
Summary of inversion analysis with our global aerosol model Preliminary results of our inversion analyses adjusted the emission flux of 137 Cs to 13 24 PBq, which are between JAEA (9.1 PBq) and NILU (36.6 PBq, Stohl et al. 2012). The estimated emission flux is highly sensitive to the selection of the first guess, because of the under constraint of observations.
To reduce uncertainty, We should further investigate for the sensitivities in microphysical and deposition processes of aerosols, and use of observed deposition fluxes of 137 Cs. Ex. Aerosol size distribution: Stohl et al. (2012) : Lognormal: 0.4μm (aerodynamic mean), σ = 0.3 Takemura et al. (2011): 10μm
Simulation and estimation with a regional transport model
Questions In this model, we are trying to formulate the behavior of atmospheric radioactive materials in detail, and examine which processes are important. To reproduce the deposition flux of radioactive materials, we have to reproduce the precipitation amount and other processes in Dry deposition Wet deposition by rain Wet deposition by snow How do the chemical composition, size distribution, and surface conditions affect the simulated results?
Behavior of the radioactive materials in the atmosphere 131 I: 10 15 Bq/h 1g/h (10 6 m 3 ) 1μg/m 3 /h ng~pg/m 3 Hot vapor Environment species μg/m 3 100 Bq/m 3 1 pg/m 3 (D=100nm) 10 3 particles /m 3 +Pollen From SCOPE50, review paper of Chernobyl accident, Warner and Harrison (1993)
MANAL NHM PM/r (Δx=4km) (Passive tracers Model for radioactivity, off line coupled with NHM) Gas Condensation/ Evaporation Aerosols ATK PRI Dissolution CCN activation CLD Hydrometeors RNW Dry deposition Coagulation ACM SS DU POL IN activation Coalescence Cloud Microphysics ICE SNW GRW Wet deposition (Rain, Snow) Sub micron (~0.1μm) Super micron (~1μm) Super large (~10μm) ATK (Aitken), PRI (primary nuclei), ACM (accumulation mode) DU (soil particle), SS (Sea salt particle) POL (pollen)
Simulated distributions of radioactive materials in aerosol for 15 and 20 March Observed I 131 in Tsukuba: 70Bq/m 3 Observed Cs 137 in Tsukuba: 40Bq/m 3
Uncertainties in wet removal processes Accumulated deposition of Cs 137 by rain (kbq/m 2 Accumulated deposition of Cs 137 by snow(kbq/m 2 ) Aircraft monitoring of deposited Cs 137 (MEXT)
Size distribution and dry deposition velocity Dg < 1μm (ATK+PRI+ACM) Dg > 1μm (DU+SS) Dg > 10μm (POL) 100 10 0.1 1000 10~100 Re emission 1 Averaged dry deposition velocity during 3/11 3/22 : Function of wind speed and roughness 0.2~1 1~10 5~10 Ground level concentration: fine particles dominate CsI, coarse particles are one order smaller, pollens are two order smaller. Dry deposition velocity is one order larger over land than over ocean, coarse particle are similar or one order larger than fine particles, pollens are one order larger.
How do the dry deposition velocity of gas phase and aerosol phase iodine? 0.5 2 (3/11 3/22 average) ACM SO 2 1 1 0.02 0.01 I 2 10 7 CH 3 I Dry deposition velocity of I 2 and CH 3 I are 2 order smaller than those of SO2 and submicron particles.
Grid average Effects of land surface types on the deposition velocity (3/11 3/22) Turbulence Quasi laminar I 2 I 2 Bare ground Mixed forest I 2 Deposition velocity of Iodine gas on water surface are 5 times faster than mixed forest, and 10 times faster than bare ground. Ground Tree Water water For submicron particles, deposition velocities of mixed forest and water surface are about 2 times faster than that of bare ground.
Inverse analysis with the regional model Observation We collected 49 sites observation data in Japan with simple quality control system. Model MRI PM by Kajino et al., Prior Information We used estimation flux by Chino et al. Preliminary estimated emission fluxes.
Estimation from the ocean deposition
Observed Cesium in the Pacific surface water (M. Aoyama, MRI/JMA)
Multi model calculation: Atmospheric transport models calculation for deposition behaviour over the North Pacific Ocean
Ocean transport model calculation for radiocaesium activity in the ocean
Estimated Cs 137 in the Ocean Environment Estimated atmospheric input to Pacific 13.8 16.1 PBq Direct discharge form the power plant 3.5 ± 0.7 PBq Cs 137 in Pacific Ocean before the accident, mainly due to nuclear weapon test 69 PBq
Estimated total 137 Cs emission flux Total Flux This study 13 24 PBq (3/10 4/20) JAEA (Chino et al., revised 2011) 9.1 PBq (3/10 3/31) Stohl et al. (2012) 36.6 PBq (20.1 53.1) (3/10 4/20) Winiarek et al. (2012) 10 19 PBq MELCOR analysis (Gauntt et al.) 16.4 PBq From Stohl et al. (2012) IRSN 10 PBq ZAMG 66.6 PBq From Stohl et al. (2012) Aoyama et al. (ms. in preparation) 15.2 20.4 PBq From obs. and numerical model analysis
Summary and Future Plans We are constructing an inverse modeling system to estimate emission flux by Fukushima power plant. Preliminary emission flux was obtained. We have a plan to estimate emission flux in higher time resolution and longer period. We have a plan to provide our estimated flux to global/regional model to understand transport and deposition processes. It is recommended that we need inversion intercomparison to avoid estimation error by using only one model.
Future plans Data assimilation of radioactive materials using Ensemble Kalman Filter Data assimilation with radiation dose rate
This is the end of the presentation. Thank you very much! Grazie mille!
Comparison of the time series of the emission of Cesium 137